US8566086B2 - System for adaptive enhancement of speech signals - Google Patents
System for adaptive enhancement of speech signals Download PDFInfo
- Publication number
- US8566086B2 US8566086B2 US11/167,955 US16795505A US8566086B2 US 8566086 B2 US8566086 B2 US 8566086B2 US 16795505 A US16795505 A US 16795505A US 8566086 B2 US8566086 B2 US 8566086B2
- Authority
- US
- United States
- Prior art keywords
- spectral shape
- speech
- speech signal
- correction factor
- background noise
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active, expires
Links
- 230000003044 adaptive effect Effects 0.000 title description 3
- 230000003595 spectral effect Effects 0.000 claims abstract description 275
- 238000001228 spectrum Methods 0.000 claims abstract description 120
- 238000012937 correction Methods 0.000 claims abstract description 108
- 238000000034 method Methods 0.000 claims abstract description 46
- 230000004044 response Effects 0.000 claims abstract description 41
- 230000002708 enhancing effect Effects 0.000 claims abstract description 18
- 239000000872 buffer Substances 0.000 claims description 59
- 230000007774 longterm Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims 1
- 238000001914 filtration Methods 0.000 abstract description 3
- 238000001514 detection method Methods 0.000 description 10
- 238000012935 Averaging Methods 0.000 description 6
- 238000010606 normalization Methods 0.000 description 6
- 230000006978 adaptation Effects 0.000 description 4
- 230000001629 suppression Effects 0.000 description 4
- 238000003786 synthesis reaction Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 238000012546 transfer Methods 0.000 description 3
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000001627 detrimental effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 239000002085 irritant Substances 0.000 description 1
- 231100000021 irritant Toxicity 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
- G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
- G10L15/065—Adaptation
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/20—Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
Definitions
- the present invention relates to a system and method for adaptively enhancing the frequency response of a speech signal in real-time.
- a speech signal received at a microphone and input to an audio application may be adversely impacted by slowly varying, or time-invariant acoustical or electrical characteristics of the acoustical environment or the electrical audio path.
- the in-car acoustics or microphone characteristics can have a significant detrimental impact on the sound quality or intelligibility of a speech signal transmitted to a remote party.
- Adjusting the spectral shape of a received speech signal can significantly improve the quality of the speech signal.
- the spectral shape of a speech signal may be adjusted to compensate for excessive background noise.
- the overall sound quality or intelligibility of the signal can be significantly improved.
- the ideal spectral shape for a handsfree telephone system may be significantly different from the ideal spectral shape for a speech recognition system. In the first case, it is desirable to improve both sound quality and intelligibility, in the second it may be more desirable to improve the intelligibility of the speech signal with little or no regard to the actual sound quality.
- FIG. 1 shows two examples of desirable frequency responses for two different applications.
- the first frequency response curve 10 represents a spectral shape intended to provide optimal speech quality in an environment with a high a signal-to-noise ratio (SNR).
- the second frequency response curve 12 shows a spectral shape intended to provide optimal speech intelligibility in a low signal-to-noise environment.
- FIG. 1 also shows VDA (Verband der Automobilindustrie) and ITU (International Telecommunications Union) upper and lower spectral limits 14 , 16 for the frequency response in hands-free telephony systems. In some cases it may also be desirable to adjust the spectral shape of a received speech signal to conform with the VDA and ITU limits for speech frequency response.
- a speech signal recorded by a microphone and input to an audio application will have an actual spectral shape significantly different from the ideal spectral shape for the application. Accordingly, adjusting the spectrum of the speech signal to more closely conform to the ideal spectral shape is desirable.
- a system and method for performing such an adjustment, or normalization must be capable of taking into account the acoustic transfer function characteristics of the environment in which the speech signal is recorded, and the frequency response of the electrical audio path. Furthermore, such a system and method must also take into account acoustic and electrical changes that may occur in the systems.
- a system and method for adaptively enhancing speech signals are provided.
- the system and method of the invention affectively normalize the spectrum of an input speech signal toward a target spectral shape, or ideal frequency response.
- the target spectral shape may be selected based on the application for which the speech signal is intended. For example, a desired spectral shape for a speech signal destined to be transmitted via a handsfree telephone in an automobile may be significantly different from the desired spectral shape of a speech signal which is to be input into a speech recognition system.
- an average speech spectral shape estimate is calculated based on speech signals received over time.
- the average speech spectral shape estimate may be calculated using first order IIR filtering or “leaky integrators.”
- the average speech spectral shape estimate adapts to changes in the acoustic characteristics of the voice path or any changes in the electrical audio path that may affect the frequency response of the system.
- the spectral correction factor may be determined by comparing the average speech spectral shape estimate to the desired or target spectral shape.
- the spectral correction factor represents on average, the differences in the time-averaged spectral energy of received speech signals and the desired frequency response.
- the spectral correction factor may be added to the spectrum of the input speech signal in order to normalize, or adjust, the spectrum of the input speech signal toward the desired spectral shape.
- an embodiment of a method of normalizing a speech signal will include determining the average spectral shape of the input speech. The method further includes comparing the average spectral shape of the input speech to the target spectral shape. Differences between the target spectral shape and the average spectral shape of speech that has been received over time may be used to correct the spectrum of the input speech signal. The corrected spectrum of the speech signal will more closely match the desired spectral shape for the particular application for which the speech signal is intended.
- the frequency response of the speech signal is enhanced in real-time.
- a frequency sub-band analysis is performed on successive overlapping windowed buffers of the input speech signal.
- the results of the frequency sub-band analysis of each successive windowed buffer are used to calculate an average speech spectral shape estimate.
- the average speech spectral shape estimate is then subtracted from the desired target spectral shape.
- the difference between the target spectral shape and the average speech spectral shape estimate form a spectral shape correction factor.
- the spectral shape correction factor may then be added to the spectrum corresponding to the windowed buffer of the input speech signal. Corrected spectra from successive windowed buffers may then be re-synthesized into an enhanced or normalized voice signal.
- Another embodiment enhances the frequency response of a speech signal by adjusting the spectral shape of the background noise of a received speech signal.
- This embodiment includes performing a frequency sub-band analysis on successive overlapping windowed buffers of a speech signal. A background noise estimate is generated based on the received signal. Next, a background noise spectral shape correction factor is calculated by subtracting the background noise estimate from a target background noise spectral shape. The background noise spectral shape correction factor is then added to a spectrum corresponding to one of the successive overlapping windowed buffers.
- a method of enhancing a frequency response of a speech signal also includes performing a frequency sub-band analysis on successive overlapping windowed buffers of a speech signal.
- An average speech spectral shape estimate is calculated based on the frequency sub-band analysis of successive overlapping windowed buffers.
- a speech spectral shape correction factor is calculated according to the difference between the average speech spectral shape estimate and a target speech spectral shape.
- the background noise included in the received signal is estimated and a background noise spectral shape correction factor is calculated corresponding to differences between the background noise estimate and a target background noise spectral shape.
- the speech spectral shape correction factor and the background noise spectral shape correction factor are combined to form an overall spectral shape correction factor.
- the overall spectral shape correction factor is then applied to a spectrum corresponding to one of the successive overlapping windowed buffers of the received speech signal.
- a system for enhancing the frequency response of a speech signal includes a microphone for receiving the speech signal.
- An A/D converter converts the speech signal into a digital audio signal which is input to a processor.
- the processor is adapted to determine an average speech spectral shape estimate of the speech recorded by the microphone.
- the processor compares the average speech spectral shape estimate to a target spectral shape.
- the processor then adjusts the spectral shape of the input speech signal based on differences between the average speech spectral shape estimate and the target spectral shape.
- the processor outputs a normalized speech signal having an enhanced frequency response which is closer to the ideal frequency response for the particular application for which the speech signal is intended.
- the processor may also be adapted to determine a background noise spectral shape estimate of a received signal.
- the processor may then compare the background noise spectral shape estimate with a target background noise spectral shape.
- the processor may then adjust the spectral shape of the input speech signal based on differences between the background noise spectral shape estimate and the target background noise spectral shape.
- the processor may then output a normalized speech signal having an enhanced frequency response that has a background noise spectral shape which is closer to the desired background noise spectral shape
- FIG. 1 is a dB v. frequency plot showing two examples of ideal frequency responses, or target speech spectral shapes, for two different applications.
- FIG. 2 is a flow chart illustrating a method of enhancing the frequency response of a speech signal.
- FIG. 3 shows a time domain speech signal and a plurality of overlapping windowed buffers.
- FIG. 4 is a dB v. frequency plot of the spectrum of the speech signal of FIG. 3 corresponding to one of the windowed buffers.
- FIG. 5 is a dB v. frequency plot of a frequency-compressed version of the spectrum shown in FIG. 4 , along with a background noise estimate.
- FIG. 6 is a dB v. frequency plot of the compressed spectrum of FIG. 5 with the background noise subtracted (i.e. SNR) and a threshold value representing a signal level 10 dB above the background noise.
- SNR background noise subtracted
- FIG. 7 is a dB v. frequency plot of an average speech spectral shape estimate and a target spectral shape.
- FIG. 8 is a dB v. frequency plot of a spectral correction factor derived by subtracting the average speech spectral shape estimate of FIG. 7 from the target spectral shape also shown in FIG. 7 .
- FIG. 9 is a dB v. frequency plot showing both the original spectrum of the speech signal (i.e. from FIG. 4 ) corresponding to one of the windowed buffers, and the enhanced or normalized spectrum of the speech signal.
- FIG. 10 is a spectrogram—time v. frequency v. dB (in shades of grey)—of an input speech signal.
- FIG. 11 is a spectrogram showing the adaptation of the average speech spectral shape estimate over time.
- FIG. 12 is a flowchart illustrating an alternative embodiment of a method of enhancing the frequency response of a speech signal.
- FIG. 13 is a block diagram of a system for enhancing the frequency response of the speech signal according to the invention
- a system and methods for adaptively enhancing the frequency response of a speech signal in real-time are provided.
- the system and methods compensate for the spectral effects of any slowly varying or time-invariant acoustical or electrical characteristics of the audio and electrical paths, including for example, room acoustics, microphone frequency response, and other factors such as background noise, and the like.
- the system and methods include provisions for receiving an input speech signal, calculating an average speech spectral shape estimate and a background noise estimate, calculating a correction factor for adjusting the average speech spectral shape estimate to conform to a target speech spectral shape, or adjusting the spectral shape of the background noise, and applying the correction factor to spectra of successive windowed buffers of the input speech signal in order to arrive at a desired target frequency response specifically adapted for a particular application.
- the corrected spectra corresponding to successive windowed buffers may be re-synthesized into an enhanced speech signal more suitable to the application for which the speech signal is intended.
- FIG. 2 shows a flow chart 100 of a method for adaptively enhancing the frequency response of a speech signal according to the invention.
- An input speech signal is received at 102 .
- the signal may or may not include speech content.
- a frequency sub-band analysis is performed on the input signal at 104 .
- the frequency sub-band analysis results in a compressed dB spectrum of the input signal.
- the compressed dB spectrum is used to create an average speech spectral shape estimate, which in turn is used to calculate a speech spectral shape correction factor which may be added back to the spectrum of the input signal to create an enhanced speech signal having approximately the desired spectral shape.
- the compressed dB spectrum generated by the sub-band analysis 104 is applied to SNR estimation and voice detection 106 .
- the SNR estimation and voice detection 106 attempt to identify which frequency bins of the compressed dB spectrum of the input signal contain speech content.
- the dB values of the frequency bands which are found to have a high SNR and contain speech content are used to update an average speech spectral shape estimate at 108 .
- the average speech spectral shape estimate is the running average of the spectral shape of speech received over time. Because the average speech spectral shape is a running average, it adapts slowly and provides a reasonably accurate estimate of the spectral shape of the speech content of the received input signal.
- the average speech spectral shape accounts for the slowly varying or time invariant frequency response characteristics of the system, including the acoustical transfer function characteristics of the environment, the electro-acoustic characteristics of the microphone, and the like.
- the average speech spectral shape estimate is compared to a target speech spectral shape provided at 112 .
- the target speech spectral shape may represent the ideal frequency response for a particular application, such as a handsfree telephone system or a voice recognition system. Differences between the average speech spectral shape estimate and the target speech spectral shape represent the amount by which the average spectrum of the input speech signal must be adjusted in order to achieve the desired spectral shape.
- a speech spectral shape correction factor is determined by subtracting the average speech spectral shape from the target speech spectral shape. The speech spectral shape correction factor may then be added back to the compressed dB spectrum of the original signal received at 102 .
- Background noise suppression 110 may optionally be applied to the compressed dB spectrum prior to adding the correction factor, if desired. Otherwise, the speech spectral shape correction factor is applied directly to the compressed dB spectrum at 116 . A corrected or enhanced signal is re-synthesized at 118 and output at 120 .
- FIG. 3 shows an 11 kHz time-domain speech signal 130 that is to be enhanced according to the method outlined in FIG. 2 .
- a frequency sub-band analysis is performed on successive overlapping windowed buffers.
- the windowed buffers may be calculated using 256-point Hanning windows with 50% overlap. Other windowing functions, window lengths, or overlap percentage values may also be used.
- FIG. 3 shows 50% overlapped Hanning windows 132 , 134 , 136 , 138 , 140 , and 142 .
- the frequency sub-band analysis is performed on each successive windowed buffer.
- the results of the frequency sub-band analysis from each windowed buffer contributes to the average speech spectral shape estimate.
- the analysis of a single windowed buffer 134 will be described, with the understanding that the analysis of all other windowed buffers proceeds in a like manner.
- a frequency spectrum is obtained for the portion of the signal 130 within the windowed buffer 134 .
- Frequency spectral information may be obtained by various methods such as fast Fourier transform (FFT), wavelet filter banks, polyphase filter banks, and other known algorithms.
- FFT fast Fourier transform
- a complex spectrum may be obtained using a 256-point FFT.
- Power_Spec is the power spectrum
- f is the frequency bin index
- FIG. 4 shows a dB spectrum 144 of the portion of the input signal contained within windowed buffer 134 .
- the dB spectrum 144 is the result of a 256 point FFT.
- the dB spectrum 144 includes a number of sharp peaks and valleys due to the harmonic content of a voiced speech segment (e.g. a vowel sound).
- the general shape of the spectral envelope may be analyzed by compressing the dB spectrum 144 into a spectrum having coarser frequency resolution. Frequency compression may be accomplished by calculating a weighted average across given frequency regions.
- the compressed spectrum may have a linear frequency scale, or the compressed spectrum may have a non-linear frequency scale such as a Bark, Mel, or other non-linear scale depending and the compression technique applied.
- the frequency sub-bands of the compressed spectrum may exhibit, for example, a frequency resolution of 86 to 172 Hz per compressed sub-band. For an 11 kHz input signal and a 256-point FFT, this corresponds to calculating the average power of the non-compressed spectrum across every two to four uncompressed frequency bins, respectively.
- a compressed spectrum 156 based on the uncompressed spectrum 144 of FIG. 4 is shown in FIG. 5 .
- the compressed spectrum 156 maintains the general shape of the uncompressed spectrum 144 .
- the compressed spectrum 156 represents the output of the frequency sub-band analysis 104 .
- a separate compressed spectrum is generated for each successive overlapping windowed buffer. Each contributes to the calculation of the speech spectral shape estimate.
- the average speech spectral shape estimate as updated by the frequency sub-band analysis of each successive windowed buffer, is used to calculate the speech spectral shape correction factor for the spectrum of the corresponding windowed buffer.
- the correction factor is added back to the compressed dB spectrum of the corresponding windowed buffer, to normalize the spectrum to the desired target spectral shape.
- the compressed dB spectrum generated during the frequency sub-band analysis is input to SNR estimation and voice detection 106 .
- the purpose of SNR estimation and voice detection 106 is to determine which frequency bands of the compressed dB signal have a strong signal-to-noise ratio (SNR) and are likely to contain speech. Only those frequency sub-bands of the compressed dB signal having both a high SNR and which are likely to contain speech are used to update the average speech spectral shape estimate. Those frequency bands having weak SNR or which likely do not contain speech do not contribute to the calculation of the average speech spectral shape estimate.
- SNR signal-to-noise ratio
- FIG. 5 includes a background noise estimate 158 derived using a minimum statistics technique.
- An estimate of the SNR at each frequency sub-band may be obtained by subtracting the background noise estimate 158 from the compressed dB spectrum 156 .
- FIG. 6 shows the SNR 160 that results from subtracting the noise estimate 158 from the compressed dB spectrum 156 of FIG. 5 .
- the noise estimate 158 is not the true background noise. It is just an estimate of the noise likely to be contained in the compressed dB spectrum 156 .
- the actual noise in any given frequency sub-band may be greater or less than the levels shown in the background noise estimate 158 . Thus, signal levels that are near the noise estimate are less reliable. Accordingly, a threshold value may be established such that only frequency sub-bands having a signal level above the noise estimate by an amount at least equal to the threshold value contribute to the average speech spectral shape estimate. Such a threshold is illustrated in FIG. 6 .
- the 10 dB threshold 162 represents a signal level 10 dB above the background noise estimate 158 .
- the portions of the compressed dB spectrum 160 that are above the 10 dB threshold 162 represent those portions of the original compressed dB spectrum 156 that are more than 10 dB above the background noise estimate 158 . Only those frequency sub-bands in which the compressed dB spectrum 160 is above the 10 dB threshold will contribute to the average speech spectral shape estimate.
- Threshold values other than 10 dB may be used.
- the threshold value will be in the range between 5 to 15 dB.
- the threshold need not be constant.
- the threshold value may vary from one frequency sub-band to the next, depending upon the expected noise characteristics of the system. For example, in automotive applications, the threshold could be set higher for lower frequency sub-bands where significant background noise energies reside.
- An average speech spectral shape estimate is created for each frequency sub-band of the compressed spectrum.
- the compressed spectrum for each successive overlapping windowed buffer contributes to the computation of the average speech spectral shape estimate.
- the average speech spectral shape estimate for each individual frequency sub-band is updated only when the individual frequency sub-band has a high SNR and contains speech.
- Spec_Curr ⁇ _n ⁇ ( f ) Spec_Curr ⁇ ( f ) - 1 N ⁇ ( ⁇ f N ⁇ Spec_Curr ⁇ ( f ) ) ( 2 )
- Spec_Curr is the current dB compressed spectrum
- Spec_Curr_n is the current dB compressed spectrum after overall level normalization across frequency sub-band.sf. Normalization according to equation (2) will ensure that adaptation of the average speech spectral shape will not be biased by the overall amplitude of the speech signal.
- Other level normalization techniques such as weighted averaging, frequency-dependent averaging, SNR-dependent averaging or other normalization techniques may also be used.
- the average speech spectral shape may be adapted according to a leaky-integrator algorithm, a first order IIR filter, or some other adaptive filtering or weighted averaging algorithm.
- Spec_Avg is the average speech spectral shape estimate.
- Adapt_Rate is a parameter that controls the speed of adaptation. Adapt_Rate must be >1.
- An appropriate value for AdaptTimeConstant may be between approximately 2 and 6 seconds.
- the average speech spectral shape estimate provides an estimate of the spectral shape of speech input to the system over time.
- the manner in which the average speech spectral shape estimate is generated takes into account slowly changing or time invariant acoustical characteristics of the environment, the frequency response of the microphone, the position of the microphone relative to the person speaking, and other factors that will influence the frequency response of the system.
- correction factor for each frequency sub-band may be added to the average speech spectral shape estimate in order to normalize the average speech spectral shape estimate to a desired target spectral shape.
- These same correction factors may then be applied to the individual spectra of each successive windowed buffer to enhance the input speech signal.
- the correction factors may be applied directly to the compressed dB spectra of each windowed buffer (e.g. spectrum 156 from FIG. 5 ), or they may be extrapolated and applied to the non-compressed dB spectra of the windowed buffers (e.g. spectrum 144 from FIG. 4 ).
- An average speech spectral shape estimate 166 is shown in FIG. 7 .
- a target spectral shape 168 is also shown.
- the target spectral shape may correspond to the optimum frequency response of a hands-free telephone system in an automobile, or the target spectral shape may represent the optimum speech frequency response for providing highly intelligible speech signals to a speech recognition system, or some other application.
- the target spectral shape represents the optimum frequency response toward which the dB spectrum 166 of the actual input signal is to be adjusted.
- a spectral correction factor for every sub-band of the average speech spectral shape estimate 166 may be calculated by subtracting the target spectral shape 168 from the average speech spectral shape estimate 166 .
- This difference represents the amount that must be added to or subtracted from the average speech spectral shape estimate 166 in order for the shape of the average speech spectral shape estimate 166 to exactly match the target speech spectral shape 168 .
- Spec_Target is the target speech spectral shape
- Spec_Corr is the dB spectral correction factor
- the overall level of the spectral correction values may be normalized according to:
- Spec_Corr ⁇ ( f ) Spec_Corr ⁇ ( f ) - 1 N ⁇ ( ⁇ f N ⁇ Spec_Corr ⁇ ( f ) ) ( 7 )
- This will allow for correction of the speech spectral shape without significantly modifying the overall amplitude or loudness of the speech signal.
- Other normalizing techniques such as weighted averaging or frequency-dependent averaging, or other techniques may be used.
- FIG. 8 shows the correction factor 170 calculated by subtracting the average speech spectral shape 166 from the target spectral shape 168 , as shown in equation 6, and level normalization according to equation 7.
- the correction factor 170 may by applied to the spectra of each successive windowed buffer of the input speech signal.
- the correction factor values determined above are determined for each frequency sub-band of the compressed average speech spectral shape estimate spectrum. Before being applied to the spectrum corresponding to the current windowed buffer, i.e.
- the correction values may be extrapolated to estimate correction values for all of the frequency bins of the uncompressed FFT dB spectrum. This may be performed using simple linear interpolation or cubic spline interpolation, or some other algorithm.
- the spectrum of the corresponding windowed buffer 134 may then be corrected by adding the expanded correction values (in units of dB) to the uncompressed spectrum of the input signal corresponding to the windowed buffer 134 .
- the corrected spectrum 172 corresponding to windowed buffer 134 is shown in FIG. 9 along with the original spectrum 144 .
- the spectrum of a windowed buffer may be transformed back into the time domain.
- This requires converting the corrected dB spectrum 176 into an amplitude spectrum, and transforming the amplitude spectrum back to the time domain by performing a 256 point inverse FFT, or other inverse transform from the frequency domain back into the time domain.
- the time domain signal that results from the inverse FFT or other transform constitutes an enhanced speech signal corresponding to the windowed buffer 134 .
- the enhanced speech signal will have an average spectral shape that more closely resembles the target spectral shape.
- Enhanced speech signals are re-synthesized for each windowed buffer, and are overlapped and added together in the time domain.
- the result is a re-synthesized time domain speech signal that substantially maintains a desired spectral shape over time, taking into account slowly changing characteristics of the system's transfer function.
- the result is an enhanced voice signal that better serves the particular application for which it is intended, be it a speech recognition system, a hands free telephone system, or some other application.
- FIGS. 10 and 11 show spectrogram plots which illustrate the adaptive qualities of the method just described. Both figures display plots of frequency (vertical axes) v. time (horizontal axis) v. dB (gray scale).
- the plot 180 in FIG. 10 represents the original speech signal without correction.
- the background noise frequency response rather than the speech signal frequency response.
- background noise is not a significant problem and enhancing the speech signal spectral shape is most appropriate.
- background noise having tonal qualities has been found to be more annoying to listeners than broadband noise.
- the quality and intelligibility of a speech signal is enhanced by targeting and shaping the background noise spectrum of the received speech signal as opposed to enhancing the spectrum of the speech components themselves.
- a flow chart 300 embodying this alternative is shown in FIG. 12 .
- the flow chart 300 in FIG. 12 has many similarities to the flow chart 100 shown in FIG. 2 .
- the method for adaptively enhancing the frequency response of the speech signal embodied in the flow chart 100 is substantially repeated in flow chart 300 .
- the apply speech spectral shape correction factor 116 and signal re-synthesis 118 of FIG. 2 likewise have parallels in FIG. 12 , namely apply spectral correction factor 316 and signal re-synthesis 318 .
- the apply spectral shape correction factor of 316 and signal re-synthesis 318 functions perform substantially the same functions as their counterparts in the earlier embodiment, they perform these functions on somewhat different input.
- the output of the determine speech spectral shape correction factor 314 is a speech spectral shape correction factor that may be added to the spectrum of the input signal 302 to correct or normalize the spectral shape of the input signal 302 much like the output of the corresponding determine speech spectral shape correction factor function 114 of flowchart 100 .
- the speech spectral shape correction factor is applied directly to the spectrum of the input signal (optionally after background noise suppression has been applied to the input speech signal spectrum)
- the speech spectral shape correction factor determined at 314 is input to determine final spectral correction factor 326 .
- Determine final spectral correction factor 326 also receives input from determine background noise spectral shape correction factor 326 .
- a final spectral correction factor is determined based on both a speech spectral shape correction factor and a background noise spectral shape correction factor.
- an input speech signal is received at 302 .
- the input speech signal may include background noise.
- the input speech signal is subjected to a frequency sub-band. analysis at 304 .
- the result of the frequency sub-band. analysis is a compressed dB scale spectrum representing the input speech signal.
- the compressed dB speech signal spectrum is input to SNR estimation and voice detection 306 .
- SNR estimation and voice detection 306 produces a background noise estimate 322 which is input to determine background noise spectral shape correction factor 326 .
- the background noise estimate 322 provides an estimate in dB of the background noise across each frequency bin of the compressed dB spectrum of the input speech signal 302 .
- the background noise estimate 312 may include unwanted peaks or other characteristics at various frequencies which are detrimental to the speech signal sound quality and intelligibility. Therefore, it is desirable to smooth the background noise estimate or otherwise shape the background noise estimate to conform to a desired target background noise spectral shape 324 .
- the target background noise spectral shape is input to determine background noise spectral shape correction factor 326 .
- the difference between background noise estimate 322 and the target background noise spectral shape represents the amount by which the background noise estimate must be adjusted in order to conform to the shape of the target background noise spectral shape.
- the determine background noise spectral correction factor 326 calculates a background noise spectral correction factor by subtracting the target speech spectral shape from the background noise estimate across all frequency bins of the compressed dB spectrum of the input signal.
- the background noise spectral shape correction factor may be added directly to the compressed dB spectrum of the input speech signal 302 in order to shape the frequency spectrum of the background noise included in the input speech signal 302 .
- both the speech spectral shape correction factor and the background noise spectral shape correction factor contribute to a final spectral shape correction factor.
- the final spectral shape correction factor is then added to the compressed db spectrum of the input speech signal 302 .
- the output of the determine speech spectral shape correction factor 314 and the output from the determine background noise spectral shape correction factor 328 are both input to the determine final spectral shape correction factor 328 .
- the speech spectral shape correction factor and the background noise spectral shape correction factor contribute to the final spectral shape correction factor in an inversely proportional manner according to the formula:
- Final_corr( f ) a *Speech_Corr( f )+(1 ⁇ a )*Noise_Corr( f ) (10) where
- Speech_Corr(f) Speech Spectral Shape Correction Factor
- Noise_Corr(f) Background Noise Spectral Shape Correction Factor
- FIG. 13 shows a block diagram of such a system 200 .
- the system includes a microphone 202 an A/D converter 204 and a signal processor 206 .
- the microphone 202 captures an input signal.
- the A/D converter samples the analog signal from the microphone and provides a digital signal representing the speech and background noise received by the microphone to the signal processor 206 .
- the processor 206 includes instructions for performing all of the steps described above on the input sign a captured by microphone 202 .
- the processor performs frequency sub-band analysis, SNR estimation and voice detection on the input signal.
- the processor creates and updates an average speech spectral shape estimate for every windowed buffer of the input speech signal, and stores a target speech spectral shape. For each windowed buffer the processor calculates a spectral correction factor for matching the average speech spectral shape estimate to the target speech spectral shape.
- the processor may also determine a background noise spectral shape correction factor based on a background noise estimate and a stored target background noise spectral shape.
- the processor may apply either the speech spectral shape correction factor or the background noise spectral correction factor to the spectra of each windowed buffer or the processor may apply a final correction factor comprising a composite of the speech spectral shape correction factor and the background noise spectral shape correction factor.
- the processor then converts the spectra back into the time domain, and re-synthesizes an enhanced output signal 208 .
- the output signal 208 may then in turn be applied as an input to another system that employs the enhanced speech signal.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Acoustics & Sound (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Noise Elimination (AREA)
- Telephone Function (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Telephonic Communication Services (AREA)
Abstract
Description
Power_Spec(f)=|Complex_Spec(f)|2 (1)
where
where Spec_Curr is the current dB compressed spectrum and Spec_Curr_n is the current dB compressed spectrum after overall level normalization across frequency sub-band.sf. Normalization according to equation (2) will ensure that adaptation of the average speech spectral shape will not be biased by the overall amplitude of the speech signal. Other level normalization techniques such as weighted averaging, frequency-dependent averaging, SNR-dependent averaging or other normalization techniques may also be used.
Spec_Avg(f)=(Spec_Avg(f)*(Adapt_Rate−1)+Spec_Curr— n(f)/Adapt_Rate (3)
where
Adapt_Rate=AdaptTimeConstant/SecPerBuffer (4)
SecPerBuffer=(FFT Size-BufferOverlap)/Sample Rate (5)
Spec_Corr(f)=Spec_Target(f)−Spec_Avg(f) (6)
where
This will allow for correction of the speech spectral shape without significantly modifying the overall amplitude or loudness of the speech signal. Other normalizing techniques, such as weighted averaging or frequency-dependent averaging, or other techniques may be used.
Spec_Corr(f)=Max(Spec_Corr(f), −Core_dB_Limit) (8)
Spec_Corr(f)=Min(Spec_Corr(f), Corr_dB_Limit) (9)
Typical values for Corr_dB_Limit may be in the range between 5 and 15 dB.
Final_corr(f)=a*Speech_Corr(f)+(1−a)*Noise_Corr(f) (10)
where
-
- If the long term SNR is high a→1
- If the long term SNR is low a→0
Thus, in high SNR conditions the speech spectral shape correction factor (Speech_Corr(f)) predominates, and in low SNR conditions the background noise spectral shape correction factor (Noise_Corr (f)) predominates. Once the final spectral shape correction factor has been determined, it is applied to the spectrum of the input speech signal at 316. As with the embodiment shown inFIG. 2 , the final spectral shape correction factor is added to the dB spectrum of the received speech signal output from the frequency sub-band.analysis 304. The final corrected or enhanced spectrum is then re-sysnthesized at 318. The re-synthesis process is substantially the same as that described above with regard to the embodiment depicted inFIG. 2 . The final enhanced signal is output at 320.
Claims (18)
Priority Applications (6)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/167,955 US8566086B2 (en) | 2005-06-28 | 2005-06-28 | System for adaptive enhancement of speech signals |
CA2549744A CA2549744C (en) | 2005-06-28 | 2006-06-09 | System for adaptive enhancement of speech signals |
EP06012166A EP1739657B1 (en) | 2005-06-28 | 2006-06-13 | Speech signal enhancement |
JP2006164641A JP2007011330A (en) | 2005-06-28 | 2006-06-14 | System for adaptive enhancement of speech signal |
CN2006100930862A CN1971711B (en) | 2005-06-28 | 2006-06-20 | System for adaptive enhancement of speech signals |
KR1020060055210A KR20070000987A (en) | 2005-06-28 | 2006-06-20 | System for adaptive enhancement of speech signals |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/167,955 US8566086B2 (en) | 2005-06-28 | 2005-06-28 | System for adaptive enhancement of speech signals |
Publications (2)
Publication Number | Publication Date |
---|---|
US20060293882A1 US20060293882A1 (en) | 2006-12-28 |
US8566086B2 true US8566086B2 (en) | 2013-10-22 |
Family
ID=37067632
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/167,955 Active 2030-12-17 US8566086B2 (en) | 2005-06-28 | 2005-06-28 | System for adaptive enhancement of speech signals |
Country Status (6)
Country | Link |
---|---|
US (1) | US8566086B2 (en) |
EP (1) | EP1739657B1 (en) |
JP (1) | JP2007011330A (en) |
KR (1) | KR20070000987A (en) |
CN (1) | CN1971711B (en) |
CA (1) | CA2549744C (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120115520A1 (en) * | 2009-04-22 | 2012-05-10 | Roessel Sabine | Selective Interference Rejection Combining |
US20130090922A1 (en) * | 2011-10-07 | 2013-04-11 | Pantech Co., Ltd. | Voice quality optimization system and method |
US20130297306A1 (en) * | 2012-05-04 | 2013-11-07 | Qnx Software Systems Limited | Adaptive Equalization System |
US20140278393A1 (en) * | 2013-03-12 | 2014-09-18 | Motorola Mobility Llc | Apparatus and Method for Power Efficient Signal Conditioning for a Voice Recognition System |
Families Citing this family (47)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7590523B2 (en) * | 2006-03-20 | 2009-09-15 | Mindspeed Technologies, Inc. | Speech post-processing using MDCT coefficients |
JP4757158B2 (en) * | 2006-09-20 | 2011-08-24 | 富士通株式会社 | Sound signal processing method, sound signal processing apparatus, and computer program |
US8195454B2 (en) | 2007-02-26 | 2012-06-05 | Dolby Laboratories Licensing Corporation | Speech enhancement in entertainment audio |
BRPI0808965B1 (en) * | 2007-06-19 | 2020-03-03 | Dolby Laboratories Licensing Corporation | METHOD AND APPARATUS TO MEASURE THE PERCEIVED SOUND INTENSITY OF AN AUDIO SIGN AND MEDIA READABLE BY COMPUTER |
US8625659B2 (en) * | 2008-01-10 | 2014-01-07 | Viasat, Inc. | Receiver-based frequency response estimation |
US8606573B2 (en) * | 2008-03-28 | 2013-12-10 | Alon Konchitsky | Voice recognition improved accuracy in mobile environments |
US9373339B2 (en) * | 2008-05-12 | 2016-06-21 | Broadcom Corporation | Speech intelligibility enhancement system and method |
US9197181B2 (en) * | 2008-05-12 | 2015-11-24 | Broadcom Corporation | Loudness enhancement system and method |
US8831936B2 (en) * | 2008-05-29 | 2014-09-09 | Qualcomm Incorporated | Systems, methods, apparatus, and computer program products for speech signal processing using spectral contrast enhancement |
US8538749B2 (en) * | 2008-07-18 | 2013-09-17 | Qualcomm Incorporated | Systems, methods, apparatus, and computer program products for enhanced intelligibility |
CN102150206B (en) * | 2008-10-24 | 2013-06-05 | 三菱电机株式会社 | Noise suppression device and audio decoding device |
US9202456B2 (en) | 2009-04-23 | 2015-12-01 | Qualcomm Incorporated | Systems, methods, apparatus, and computer-readable media for automatic control of active noise cancellation |
US20110015922A1 (en) * | 2009-07-20 | 2011-01-20 | Larry Joseph Kirn | Speech Intelligibility Improvement Method and Apparatus |
GB2485510B (en) * | 2009-09-15 | 2014-04-09 | Hewlett Packard Development Co | System and method for modifying an audio signal |
US9838784B2 (en) | 2009-12-02 | 2017-12-05 | Knowles Electronics, Llc | Directional audio capture |
TWI562137B (en) * | 2010-04-09 | 2016-12-11 | Dts Inc | Adaptive environmental noise compensation for audio playback |
US9053697B2 (en) | 2010-06-01 | 2015-06-09 | Qualcomm Incorporated | Systems, methods, devices, apparatus, and computer program products for audio equalization |
US8831937B2 (en) * | 2010-11-12 | 2014-09-09 | Audience, Inc. | Post-noise suppression processing to improve voice quality |
JP2012133205A (en) * | 2010-12-22 | 2012-07-12 | Sony Corp | Noise reduction device and method, and program |
US8965756B2 (en) * | 2011-03-14 | 2015-02-24 | Adobe Systems Incorporated | Automatic equalization of coloration in speech recordings |
JP5278477B2 (en) * | 2011-03-30 | 2013-09-04 | 株式会社ニコン | Signal processing apparatus, imaging apparatus, and signal processing program |
US20130282373A1 (en) * | 2012-04-23 | 2013-10-24 | Qualcomm Incorporated | Systems and methods for audio signal processing |
CN104662605A (en) * | 2012-07-25 | 2015-05-27 | 株式会社尼康 | Signal processing device, imaging device, and program |
US9210507B2 (en) | 2013-01-29 | 2015-12-08 | 2236008 Ontartio Inc. | Microphone hiss mitigation |
EP2760221A1 (en) * | 2013-01-29 | 2014-07-30 | QNX Software Systems Limited | Microphone hiss mitigation |
JP6232710B2 (en) * | 2013-02-22 | 2017-11-22 | 大日本印刷株式会社 | Sound recording device |
CN111179954B (en) * | 2013-03-04 | 2024-03-12 | 声代Evs有限公司 | Apparatus and method for reducing quantization noise in a time domain decoder |
US20140278380A1 (en) * | 2013-03-14 | 2014-09-18 | Dolby Laboratories Licensing Corporation | Spectral and Spatial Modification of Noise Captured During Teleconferencing |
US9536540B2 (en) | 2013-07-19 | 2017-01-03 | Knowles Electronics, Llc | Speech signal separation and synthesis based on auditory scene analysis and speech modeling |
WO2015059947A1 (en) * | 2013-10-22 | 2015-04-30 | 日本電気株式会社 | Speech detection device, speech detection method, and program |
US9721580B2 (en) * | 2014-03-31 | 2017-08-01 | Google Inc. | Situation dependent transient suppression |
DE112015004185T5 (en) | 2014-09-12 | 2017-06-01 | Knowles Electronics, Llc | Systems and methods for recovering speech components |
KR101610161B1 (en) * | 2014-11-26 | 2016-04-08 | 현대자동차 주식회사 | System and method for speech recognition |
CN104900237B (en) * | 2015-04-24 | 2019-07-05 | 上海聚力传媒技术有限公司 | A kind of methods, devices and systems for audio-frequency information progress noise reduction process |
CN105913854B (en) * | 2016-04-15 | 2020-10-23 | 腾讯科技(深圳)有限公司 | Voice signal cascade processing method and device |
US9820042B1 (en) | 2016-05-02 | 2017-11-14 | Knowles Electronics, Llc | Stereo separation and directional suppression with omni-directional microphones |
CN106448696A (en) * | 2016-12-20 | 2017-02-22 | 成都启英泰伦科技有限公司 | Adaptive high-pass filtering speech noise reduction method based on background noise estimation |
US10504538B2 (en) | 2017-06-01 | 2019-12-10 | Sorenson Ip Holdings, Llc | Noise reduction by application of two thresholds in each frequency band in audio signals |
CN108318883B (en) * | 2018-01-16 | 2020-04-03 | 中国科学院声学研究所 | Distance direction calculation method based on natural weighting of full-spectrum transducer |
CN109741760B (en) * | 2018-12-18 | 2020-12-22 | 科大讯飞股份有限公司 | Noise estimation method and system |
CN109979469B (en) * | 2019-04-03 | 2021-04-23 | 北京小米智能科技有限公司 | Signal processing method, apparatus and storage medium |
CN110191396B (en) * | 2019-05-24 | 2022-05-27 | 腾讯音乐娱乐科技(深圳)有限公司 | Audio processing method, device, terminal and computer readable storage medium |
WO2021021683A1 (en) * | 2019-07-30 | 2021-02-04 | Dolby Laboratories Licensing Corporation | Method and apparatus for normalizing features extracted from audio data for signal recognition or modification |
EP3944237A1 (en) * | 2020-07-21 | 2022-01-26 | EPOS Group A/S | A loudspeaker system provided with dynamic speech equalization |
CN113112990A (en) * | 2021-03-04 | 2021-07-13 | 昆明理工大学 | Language identification method of variable-duration voice based on spectrum envelope diagram |
CN113259825B (en) * | 2021-06-03 | 2021-10-08 | 南京天悦电子科技有限公司 | Low-complexity amplitude-frequency response later-period correction method for digital hearing aid |
CN114112006A (en) * | 2021-11-26 | 2022-03-01 | 中科传启(苏州)科技有限公司 | Noise monitoring method and device and electronic equipment |
Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4454609A (en) * | 1981-10-05 | 1984-06-12 | Signatron, Inc. | Speech intelligibility enhancement |
US4630304A (en) * | 1985-07-01 | 1986-12-16 | Motorola, Inc. | Automatic background noise estimator for a noise suppression system |
US5251263A (en) * | 1992-05-22 | 1993-10-05 | Andrea Electronics Corporation | Adaptive noise cancellation and speech enhancement system and apparatus therefor |
US5485522A (en) * | 1993-09-29 | 1996-01-16 | Ericsson Ge Mobile Communications, Inc. | System for adaptively reducing noise in speech signals |
US5655057A (en) * | 1993-12-27 | 1997-08-05 | Nec Corporation | Speech recognition apparatus |
US5742694A (en) * | 1996-07-12 | 1998-04-21 | Eatwell; Graham P. | Noise reduction filter |
WO1998043238A1 (en) | 1997-03-25 | 1998-10-01 | The Secretary Of State For Defence | Speech analysis system |
US5991718A (en) * | 1998-02-27 | 1999-11-23 | At&T Corp. | System and method for noise threshold adaptation for voice activity detection in nonstationary noise environments |
US6032115A (en) | 1996-09-30 | 2000-02-29 | Kabushiki Kaisha Toshiba | Apparatus and method for correcting the difference in frequency characteristics between microphones for analyzing speech and for creating a recognition dictionary |
US6035048A (en) * | 1997-06-18 | 2000-03-07 | Lucent Technologies Inc. | Method and apparatus for reducing noise in speech and audio signals |
US6275798B1 (en) * | 1998-09-16 | 2001-08-14 | Telefonaktiebolaget L M Ericsson | Speech coding with improved background noise reproduction |
US6314396B1 (en) | 1998-11-06 | 2001-11-06 | International Business Machines Corporation | Automatic gain control in a speech recognition system |
US6411927B1 (en) | 1998-09-04 | 2002-06-25 | Matsushita Electric Corporation Of America | Robust preprocessing signal equalization system and method for normalizing to a target environment |
US20050075866A1 (en) * | 2003-10-06 | 2005-04-07 | Bernard Widrow | Speech enhancement in the presence of background noise |
US20060089958A1 (en) * | 2004-10-26 | 2006-04-27 | Harman Becker Automotive Systems - Wavemakers, Inc. | Periodic signal enhancement system |
US7209567B1 (en) * | 1998-07-09 | 2007-04-24 | Purdue Research Foundation | Communication system with adaptive noise suppression |
US20080107280A1 (en) * | 2003-05-09 | 2008-05-08 | Tim Haulick | Noisy environment communication enhancement system |
US7383179B2 (en) * | 2004-09-28 | 2008-06-03 | Clarity Technologies, Inc. | Method of cascading noise reduction algorithms to avoid speech distortion |
US7454010B1 (en) * | 2004-11-03 | 2008-11-18 | Acoustic Technologies, Inc. | Noise reduction and comfort noise gain control using bark band weiner filter and linear attenuation |
-
2005
- 2005-06-28 US US11/167,955 patent/US8566086B2/en active Active
-
2006
- 2006-06-09 CA CA2549744A patent/CA2549744C/en active Active
- 2006-06-13 EP EP06012166A patent/EP1739657B1/en active Active
- 2006-06-14 JP JP2006164641A patent/JP2007011330A/en not_active Withdrawn
- 2006-06-20 CN CN2006100930862A patent/CN1971711B/en active Active
- 2006-06-20 KR KR1020060055210A patent/KR20070000987A/en not_active Application Discontinuation
Patent Citations (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4454609A (en) * | 1981-10-05 | 1984-06-12 | Signatron, Inc. | Speech intelligibility enhancement |
US4630304A (en) * | 1985-07-01 | 1986-12-16 | Motorola, Inc. | Automatic background noise estimator for a noise suppression system |
US5251263A (en) * | 1992-05-22 | 1993-10-05 | Andrea Electronics Corporation | Adaptive noise cancellation and speech enhancement system and apparatus therefor |
US5485522A (en) * | 1993-09-29 | 1996-01-16 | Ericsson Ge Mobile Communications, Inc. | System for adaptively reducing noise in speech signals |
US5655057A (en) * | 1993-12-27 | 1997-08-05 | Nec Corporation | Speech recognition apparatus |
US5742694A (en) * | 1996-07-12 | 1998-04-21 | Eatwell; Graham P. | Noise reduction filter |
US6032115A (en) | 1996-09-30 | 2000-02-29 | Kabushiki Kaisha Toshiba | Apparatus and method for correcting the difference in frequency characteristics between microphones for analyzing speech and for creating a recognition dictionary |
WO1998043238A1 (en) | 1997-03-25 | 1998-10-01 | The Secretary Of State For Defence | Speech analysis system |
US6377918B1 (en) * | 1997-03-25 | 2002-04-23 | Qinetiq Limited | Speech analysis using multiple noise compensation |
US6035048A (en) * | 1997-06-18 | 2000-03-07 | Lucent Technologies Inc. | Method and apparatus for reducing noise in speech and audio signals |
US5991718A (en) * | 1998-02-27 | 1999-11-23 | At&T Corp. | System and method for noise threshold adaptation for voice activity detection in nonstationary noise environments |
US7209567B1 (en) * | 1998-07-09 | 2007-04-24 | Purdue Research Foundation | Communication system with adaptive noise suppression |
US6411927B1 (en) | 1998-09-04 | 2002-06-25 | Matsushita Electric Corporation Of America | Robust preprocessing signal equalization system and method for normalizing to a target environment |
US6275798B1 (en) * | 1998-09-16 | 2001-08-14 | Telefonaktiebolaget L M Ericsson | Speech coding with improved background noise reproduction |
US6314396B1 (en) | 1998-11-06 | 2001-11-06 | International Business Machines Corporation | Automatic gain control in a speech recognition system |
US20080107280A1 (en) * | 2003-05-09 | 2008-05-08 | Tim Haulick | Noisy environment communication enhancement system |
US20050075866A1 (en) * | 2003-10-06 | 2005-04-07 | Bernard Widrow | Speech enhancement in the presence of background noise |
US7383179B2 (en) * | 2004-09-28 | 2008-06-03 | Clarity Technologies, Inc. | Method of cascading noise reduction algorithms to avoid speech distortion |
US20060089958A1 (en) * | 2004-10-26 | 2006-04-27 | Harman Becker Automotive Systems - Wavemakers, Inc. | Periodic signal enhancement system |
US7454010B1 (en) * | 2004-11-03 | 2008-11-18 | Acoustic Technologies, Inc. | Noise reduction and comfort noise gain control using bark band weiner filter and linear attenuation |
Non-Patent Citations (12)
Title |
---|
Acero A et al: "Acoustial Pre-Processing for Robust Spoken Language System" Proceedings of the International Conference spoken language processing (ICSLP). Kobe, Nov. 18-22, 1990 Tokyo, ASJ, JP, vol. 2, pp. 1121-1124. |
Acero A et al: "Robust Speech Recogition by Normalization of the Acoustic Space" Speech Processing 2, VLSI, Underwater Signal Processing. Toronto, May 14-17, 1991, International Conference on Acoustic, Speech & Signal Processing. ICASSP, pp. 893-896. |
Chinese Office Action (and Translation) for Chinese Application No. 200610093086.2, both created after the filing of the present application, Jun. 2, 2010. (7 pages total). |
Cohen, "On speech enhancement under signal presence uncertainty," 2001, Acoustics, Speech, and Signal Processing, 2001. Proceedings. (ICASSP '01). 2001 IEEE International Conference on , vol. 1, No., pp. 661-664 vol. 1, 2001. * |
Diethorn, "Subband Noise Reduction Methods for Speech Enhancement", In "Audio Signal Processing: For Next Generation Multimedia Communication Systems", Chapter 4, Feb. 2004, pp. 91-115. * |
European Examination Report dated May 18, 2011 for corresponding European Patent Application No. 06 012 166.2, 4 pages. |
Jensen et al, "Speech Enhancement Using a Constrained Iterative Sinusoidal Model", 2001, IEEE transactions on Speech and Audio Processing, vol. 9, No. 7, pp. 731-740. * |
Lee et al, "An improved voice activity detection algorithm employing speech enhancement preprocessing,", 2001, IEICE Trans. Fundamentals, vol. E84-A, No. 6, pp. 1401-1405, Jun. 2001. * |
Martin et al, "Optimal Recursive Smoothing of Non-Stationary Periodograms", 2001, In Proc. Int. Workshop Acoust. Echo and Noise Control (IWAENC), pp. 167-170, 2001. * |
Picone, "Signal Modeling Techniques in Speech Recognition", 1993, Proc. of the IEEE, vol. 81. No. 9, pp. 1215-1247. * |
Stern et al, "Acoustical Pre-Processing for Robust Speech Recognition", Oct. 1989, Proc. Speech and Natural Language Workshop, pp. 311-318. * |
Tilp: Single-channel Noise Reduction with Pitchadaptive Post-filtering, 2000, In Proc. EUSIPCO-2000, vol. 1, 171-174, Tampere, Finland, 2000. * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120115520A1 (en) * | 2009-04-22 | 2012-05-10 | Roessel Sabine | Selective Interference Rejection Combining |
US8798654B2 (en) * | 2009-04-22 | 2014-08-05 | Nokia Siemens Networks Oy | Selective interference rejection combining |
US20130090922A1 (en) * | 2011-10-07 | 2013-04-11 | Pantech Co., Ltd. | Voice quality optimization system and method |
US20130297306A1 (en) * | 2012-05-04 | 2013-11-07 | Qnx Software Systems Limited | Adaptive Equalization System |
US8843367B2 (en) * | 2012-05-04 | 2014-09-23 | 8758271 Canada Inc. | Adaptive equalization system |
US20140365211A1 (en) * | 2012-05-04 | 2014-12-11 | 2236008 Ontario Inc. | Adaptive equalization system |
US9099084B2 (en) * | 2012-05-04 | 2015-08-04 | 2236008 Ontario Inc. | Adaptive equalization system |
US20140278393A1 (en) * | 2013-03-12 | 2014-09-18 | Motorola Mobility Llc | Apparatus and Method for Power Efficient Signal Conditioning for a Voice Recognition System |
US20180268811A1 (en) * | 2013-03-12 | 2018-09-20 | Google Technology Holdings LLC | Apparatus and Method for Power Efficient Signal Conditioning For a Voice Recognition System |
US10909977B2 (en) * | 2013-03-12 | 2021-02-02 | Google Technology Holdings LLC | Apparatus and method for power efficient signal conditioning for a voice recognition system |
US11735175B2 (en) | 2013-03-12 | 2023-08-22 | Google Llc | Apparatus and method for power efficient signal conditioning for a voice recognition system |
Also Published As
Publication number | Publication date |
---|---|
EP1739657A3 (en) | 2007-03-07 |
EP1739657B1 (en) | 2013-01-09 |
CN1971711B (en) | 2012-07-18 |
US20060293882A1 (en) | 2006-12-28 |
EP1739657A2 (en) | 2007-01-03 |
KR20070000987A (en) | 2007-01-03 |
CA2549744C (en) | 2014-04-01 |
CA2549744A1 (en) | 2006-12-28 |
CN1971711A (en) | 2007-05-30 |
JP2007011330A (en) | 2007-01-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8566086B2 (en) | System for adaptive enhancement of speech signals | |
US8219389B2 (en) | System for improving speech intelligibility through high frequency compression | |
US8249861B2 (en) | High frequency compression integration | |
US8521530B1 (en) | System and method for enhancing a monaural audio signal | |
US7649988B2 (en) | Comfort noise generator using modified Doblinger noise estimate | |
JP4764995B2 (en) | Improve the quality of acoustic signals including noise | |
US7895036B2 (en) | System for suppressing wind noise | |
US6263307B1 (en) | Adaptive weiner filtering using line spectral frequencies | |
US5706395A (en) | Adaptive weiner filtering using a dynamic suppression factor | |
US8010355B2 (en) | Low complexity noise reduction method | |
EP2056296B1 (en) | Dynamic noise reduction | |
EP2416315B1 (en) | Noise suppression device | |
US20050240401A1 (en) | Noise suppression based on Bark band weiner filtering and modified doblinger noise estimate | |
EP1772855A1 (en) | Method for extending the spectral bandwidth of a speech signal | |
US20040057586A1 (en) | Voice enhancement system | |
WO2008121436A1 (en) | Method and apparatus for quickly detecting a presence of abrupt noise and updating a noise estimate | |
EP1081685A2 (en) | System and method for noise reduction using a single microphone | |
Itoh et al. | Environmental noise reduction based on speech/non-speech identification for hearing aids | |
RU2589298C1 (en) | Method of increasing legible and informative audio signals in the noise situation | |
Koval et al. | Broadband noise cancellation systems: new approach to working performance optimization | |
Itoh et al. | ENVIRONMENTAL NOISE REDUCTION BASED ON SPEECH/NON-SPEECH IDENTIFICATION pop | |
Avadhanulu et al. | EARLYZER: Perceptualy Motivated Robust TFR of Speech | |
JP2003517761A (en) | Method and apparatus for suppressing acoustic background noise in a communication system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: HARMAN BECKER AUTOMOTIVE SYSTEMS-WAVEMAKERS, INC., Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:HETHERINGTON, PHILLP A.;GIESBRECHT, DAVID;SIGNING DATES FROM 20050621 TO 20050623;REEL/FRAME:016734/0557 Owner name: HARMAN BECKER AUTOMOTIVE SYSTEMS-WAVEMAKERS, INC., Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:HETHERINGTON, PHILLP A.;GIESBRECHT, DAVID;REEL/FRAME:016734/0557;SIGNING DATES FROM 20050621 TO 20050623 |
|
AS | Assignment |
Owner name: QNX SOFTWARE SYSTEMS (WAVEMAKERS), INC., CANADA Free format text: CHANGE OF NAME;ASSIGNOR:HARMAN BECKER AUTOMOTIVE SYSTEMS - WAVEMAKERS, INC.;REEL/FRAME:018515/0376 Effective date: 20061101 Owner name: QNX SOFTWARE SYSTEMS (WAVEMAKERS), INC.,CANADA Free format text: CHANGE OF NAME;ASSIGNOR:HARMAN BECKER AUTOMOTIVE SYSTEMS - WAVEMAKERS, INC.;REEL/FRAME:018515/0376 Effective date: 20061101 |
|
AS | Assignment |
Owner name: JPMORGAN CHASE BANK, N.A., NEW YORK Free format text: SECURITY AGREEMENT;ASSIGNORS:HARMAN INTERNATIONAL INDUSTRIES, INCORPORATED;BECKER SERVICE-UND VERWALTUNG GMBH;CROWN AUDIO, INC.;AND OTHERS;REEL/FRAME:022659/0743 Effective date: 20090331 Owner name: JPMORGAN CHASE BANK, N.A.,NEW YORK Free format text: SECURITY AGREEMENT;ASSIGNORS:HARMAN INTERNATIONAL INDUSTRIES, INCORPORATED;BECKER SERVICE-UND VERWALTUNG GMBH;CROWN AUDIO, INC.;AND OTHERS;REEL/FRAME:022659/0743 Effective date: 20090331 |
|
AS | Assignment |
Owner name: HARMAN INTERNATIONAL INDUSTRIES, INCORPORATED,CONN Free format text: PARTIAL RELEASE OF SECURITY INTEREST;ASSIGNOR:JPMORGAN CHASE BANK, N.A., AS ADMINISTRATIVE AGENT;REEL/FRAME:024483/0045 Effective date: 20100601 Owner name: QNX SOFTWARE SYSTEMS (WAVEMAKERS), INC.,CANADA Free format text: PARTIAL RELEASE OF SECURITY INTEREST;ASSIGNOR:JPMORGAN CHASE BANK, N.A., AS ADMINISTRATIVE AGENT;REEL/FRAME:024483/0045 Effective date: 20100601 Owner name: QNX SOFTWARE SYSTEMS GMBH & CO. KG,GERMANY Free format text: PARTIAL RELEASE OF SECURITY INTEREST;ASSIGNOR:JPMORGAN CHASE BANK, N.A., AS ADMINISTRATIVE AGENT;REEL/FRAME:024483/0045 Effective date: 20100601 Owner name: HARMAN INTERNATIONAL INDUSTRIES, INCORPORATED, CON Free format text: PARTIAL RELEASE OF SECURITY INTEREST;ASSIGNOR:JPMORGAN CHASE BANK, N.A., AS ADMINISTRATIVE AGENT;REEL/FRAME:024483/0045 Effective date: 20100601 Owner name: QNX SOFTWARE SYSTEMS (WAVEMAKERS), INC., CANADA Free format text: PARTIAL RELEASE OF SECURITY INTEREST;ASSIGNOR:JPMORGAN CHASE BANK, N.A., AS ADMINISTRATIVE AGENT;REEL/FRAME:024483/0045 Effective date: 20100601 Owner name: QNX SOFTWARE SYSTEMS GMBH & CO. KG, GERMANY Free format text: PARTIAL RELEASE OF SECURITY INTEREST;ASSIGNOR:JPMORGAN CHASE BANK, N.A., AS ADMINISTRATIVE AGENT;REEL/FRAME:024483/0045 Effective date: 20100601 |
|
AS | Assignment |
Owner name: QNX SOFTWARE SYSTEMS CO., CANADA Free format text: CONFIRMATORY ASSIGNMENT;ASSIGNOR:QNX SOFTWARE SYSTEMS (WAVEMAKERS), INC.;REEL/FRAME:024659/0370 Effective date: 20100527 |
|
AS | Assignment |
Owner name: QNX SOFTWARE SYSTEMS LIMITED, CANADA Free format text: CHANGE OF NAME;ASSIGNOR:QNX SOFTWARE SYSTEMS CO.;REEL/FRAME:027768/0863 Effective date: 20120217 |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
AS | Assignment |
Owner name: 8758271 CANADA INC., ONTARIO Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:QNX SOFTWARE SYSTEMS LIMITED;REEL/FRAME:032607/0943 Effective date: 20140403 Owner name: 2236008 ONTARIO INC., ONTARIO Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:8758271 CANADA INC.;REEL/FRAME:032607/0674 Effective date: 20140403 |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
AS | Assignment |
Owner name: BLACKBERRY LIMITED, ONTARIO Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:2236008 ONTARIO INC.;REEL/FRAME:053313/0315 Effective date: 20200221 |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 8 |